최근 머신러닝은 빅데이터에 대한 분석방법으로서 학습을 통한 지능화된 문제해결 방안으로서 관심이 증가하고 있다. 본 논문은 LBSN 데이터와 머신러닝 방식을 이용하여 토지이용현황을 파악하는 분석을 시도하였다. 도시계획에 있어서 토지이용현황의 파악은 직접적인 현장 조사에 의존해 왔다. 최근 스마트폰 사용자가 증가하면서 등장하고 있는 위치기반 소셜미디어의 자료들 은 토지이용의 상황을 반영하는 빅데이터로서, 머신러닝 방법론은 이들에 대한 자동화된 분석을 할 수 있게 한다. 본 연구에서는 LBSN 자료와 머신러닝 기법을 이용하여 토지이용을 예측하는 모델을 개발하여 실제 토지이용현황 자료와의 비교분석을 수행하였다. 이러한 분석을 통해 LBSN자료를 이용한 토지이용현황의 자동화된 분석 방안에 대해 연구하였다.

For rational agricultural land use planning, it is quite necessary to get hold of land suitability precisely and to make decision on land use patterns accordingly. In the methodological viewpoint, objective and scientific evaluation techniques for land suitability classification should be supported for the systematic land use planning. As one of technical development approaches to rational land use planning, this study tried to frame a land suitability evaluation system for agricultural purposes. Evaluation unit is defined as a tract of land bounded by road, other land units and topographical features. And quantification theory was applied in the determination works of evaluation criteria. The administrative area of Namsa-myon(district), Yongin-si(city), Kyunggi-do(province) was selected for the case study. In order to check the feasibility of the evaluation system developed in the study, field check team, consisting of 2 government officers and 2 representative farmers, carried out evaluation works by observation on 148 sample land units, 10% of total 1,480 ones. Between estimated and observed results, there showed very good relationship of its multiple correlation coefficient, R=0.9467.

As a rational decision-making process of county-level area development, this study designed 3-step framework : function-giving(areal analysis) on unit planning area by decision matrix of land suitability, check of typical characteristics of each function area and formulation of its future development strategies. Two alternatives were suggested as the areal analysis method, of which one is equal ordering / valuing technique of checking indices for land suitability classfication and the other preferential weighting technique. And then, under the algorithm specially defined in this study, land suitability maps(Fig.2,3) of the case study area (Seungju-county area, Chonnam-province, Korea) were drawn from the areal analysis results. By use of land suitability classification results, unique characteristics of typical function areas were defined (on 7 types of alternative 1 , 8 types of II ) and their future development strategies were formulated in the case study area, According to the categorization criteria in this study, all the villages of the case area were classfied as a suitable type of function areas illustrated in this study.

As a initial methodological approach to rational land use planning in the county-level area, three types of land suitability classification techniques were examined from the viewpoint of their practical applicability through the case study of Seungju-gun area, Chonnam-province, Korea : major factors' criteria(method I ), principal component analysis( I ), and local monitoring( R( ). Each method has its strong and weak points as shown in Tab.-5. Therefore, when its practical application, there seem to be wide-range methodological selectivities from exclusive use of the best one to intermethodological combination of related ones In the beginning stage, intermethodological combination of all three types were tried to formulate the best solution possible. However, because of reliability problem of method R accrued from non- uniformity of evaluators'quality, only two methods( 1 , E ) were combined into a new evaluation method The applied results of the new combined method to case study area are shown in Fig.-2, 3 and 4.